MODEL / ALIBABA QWEN

Qwen3 4B

Thinking mode can increase latency and output length.

1차 출처

총 파라미터4B

추정 메모리~3.6 GBGGUF Q4

컨텍스트32K+Text

LicenseApache-2.0100+ languages and dialects

01 / 추천 용도

multilingual chat
reasoning
tools

02 / 추정 적합도

Mac Studio M5 Ultra · 96 GB여유 있음 Mac mini M6 · 24 GB여유 있음 MacBook Pro M5 Max · 64 GB여유 있음 Framework Desktop · Ryzen AI Max+ 395 · 128 GB여유 있음 Radeon AI PRO R9700 · 32 GB여유 있음

03 / 배포 환경

These are runtime examples, not validated installation recipes. Fit estimates assume quantized files; official repositories may hold full-precision weights. MLX, GGUF and TensorRT require matching conversions or engines.

OllamaEasyollama run qwen3:4b

llama.cppMediumllama-server -m /path/to/verified-model.gguf --port 9931

LM StudioEasySearch “Qwen/Qwen3-4B” → Load a Q4/INT4 build → Start Server

MLX LMMediummlx_lm.server --model Qwen/Qwen3-4B

주의

Thinking mode can increase latency and output length.